Gradient-Map-Guided Adaptive Domain Generalization for Cross Modality MRI Segmentation

This is the official code repository for Gradient-Map-Guided Adaptive Domain Generalization for Cross Modality MRI Segmentation by Bingnan Li, Zhitong Gao, Xuming He (ML4H 2023).

📜 Abstract

Cross-modal MRI segmentation is of great value for computer-aided medical diagnosis, enabling flexible data acquisition and model generaliza- tion. However, most existing methods have dif- ficulty in handling local variations in domain shift and typically require a significant amount of data for training, which hinders their us- age in practice. To address these problems, we propose a novel adaptive domain general- ization framework, which integrates a learning- free cross-domain representation based on im- age gradient maps and a class prior-informed test-time adaptation strategy for mitigating lo- cal domain shift. We validate our approach on two multi-modal MRI datasets with six cross- modal segmentation tasks. Across all the task settings, our method consistently outperforms competing approaches and shows a stable per- formance even with limited training data.

Pipeline

🚀 Guidelines

0. Platform Support

We only guarantee the correctness of the code on the following platforms:

• Linux (with cuda acceleration)
• MacOS (with MPS acceleration)

1. Create Environment

We use Python 3.11, feel free to use conda or venv to create the environment.

Once you have created the environment, install the dependencies with the following command:

pip install -r requirements.txt

2. Download the dataset

You can download the datasets used in our experiments with instructions in the following links:

• BraTS 2018
• MS-CMRSeg 2019

Once download the datasets, please place the folders into datasets with the name of BraTS2018_Raw and MS-CMRSeg2019_Raw respectively. The folder structure should be like:

BraTS2018_Raw
├── HGG
│   ├── Brats18_2013_10_1
│   ├── Brats18_2013_11_1
│   ├── ...
│   └── Brats18_TCIA08_469_1
└── LGG
    ├── Brats18_2013_0_1
    ├── Brats18_2013_15_1
    ├── ...
    └── Brats18_TCIA13_654_1

MS-CMRSeg2019_Raw
├── C0LET2_gt_for_challenge19
│   ├── C0_manual_10
│   ├── LGE_manual_35_TestData
│   └── T2_manual_10
└── C0LET2_nii45_for_challenge19
    ├── c0gt
    ├── c0t2lge
    ├── lgegt
    └── t2gt

3. Preprocess the dataset

BraTS2018

declare -a SOURCE=("t2" "flair")
declare -a TARGET=("t1" "t1ce")
for source in ${SOURCE[@]}
do
  for target in ${TARGET[@]}
  do
    python datasets/BraTS_2018.py \
            --root datasets/BraTS2018_Raw \
            --save_dir datasets/BraTS_2018 \
            --source $source \
            --target $target \
            --train_source True \
            --val_target True
  done
done

MS-CMRSeg2019

source="C0"
declare -a TARGET=("T2" "LGE")
for target in ${TARGET[@]}
do
  python datasets/MSCMRSeg2019.py \
          --root datasets/MS-CMRSeg2019_Raw \
          --save_dir datasets/MS-CMRSeg2019 \
          --source $source \
          --target $target \
          --train_source True \
          --val_target True
done

4. Train the model

cd scripts
bash train_<source_domain>.sh

To visualize the intermediate results, use the following command:

tensorboard --logdir ./saved_models/<DATASET>/<SOURCE_DOMAIN>/<EXP_NAME>

Remark:

• Only set --use_fp16 True when using NVIDIA GPU or MPS.

5. Test the model

cd scripts
bash test_<SETTING>.sh

We only provide the test scripts of C02LGE and t22t1. You can easily modify the scripts to test any settings in our paper.

Remark:

If you want to see the segmentation results and formal evaluation metrics, use the following command:

tensorboard --logdir ./val_res/<DATASET>/<SETTING>/<EXP_NAME>

Pretrained models

Dataset	Source Domain	Download	Avg_Dice(T2)	Avg_Dice(LGE)	Avg_Dice(t1)	Avg_Dice(t1ce)	Size
MS-CMRSeg2019	C0	ckpt	0.8555	0.8562	-	-	105.6M
BraTS2018	t2	ckpt	-	-	0.6813	0.6914	105.6M
BraTS2018	flair	ckpt	-	-	0.4189	0.5986	105.6M

📝 Citation

@InProceedings{pmlr-v225-li23a,
  title = 	 {Gradient-Map-Guided Adaptive Domain Generalization for Cross Modality MRI Segmentation},
  author =       {Li, Bingnan and Gao, Zhitong and He, Xuming},
  booktitle = 	 {Proceedings of the 3rd Machine Learning for Health Symposium},
  pages = 	 {292--306},
  year = 	 {2023},
  editor = 	 {Hegselmann, Stefan and Parziale, Antonio and Shanmugam, Divya and Tang, Shengpu and Asiedu, Mercy Nyamewaa and Chang, Serina and Hartvigsen, Tom and Singh, Harvineet},
  volume = 	 {225},
  series = 	 {Proceedings of Machine Learning Research},
  month = 	 {10 Dec},
  publisher =    {PMLR},
  pdf = 	 {https://proceedings.mlr.press/v225/li23a/li23a.pdf},
  url = 	 {https://proceedings.mlr.press/v225/li23a.html},
  abstract = 	 {Cross-modal MRI segmentation is of great value for computer-aided medical diagnosis, enabling flexible data acquisition and model generalization. However, most existing methods have difficulty in handling local variations in domain shift and typically require a significant amount of data for training, which hinders their usage in practice. To address these problems, we propose a novel adaptive domain generalization framework, which integrates a learning-free cross-domain representation based on image gradient maps and a class prior-informed test-time adaptation strategy for mitigating local domain shift. We validate our approach on two multi-modal MRI datasets with six cross-modal segmentation tasks. Across all the task settings, our method consistently outperforms competing approaches and shows a stable performance even with limited training data. Our Codes are available now at https://github.com/cuttle-fish-my/GM-Guided-DG .}
}

@misc{li2023gradientmapguided,
      title={Gradient-Map-Guided Adaptive Domain Generalization for Cross Modality MRI Segmentation}, 
      author={Bingnan Li and Zhitong Gao and Xuming He},
      year={2023},
      eprint={2311.09737},
      archivePrefix={arXiv},
      primaryClass={cs.CV}
}